On the Performance Evaluation of a Collaborative Swarm Intelligence Approach Particle Bee Algorithm

Abstract

Swarm intelligence (SI), an artificial intelligence (AI) approach widely used in many complex optimization problems, models the collective behavior of social systems such as honeybees and birds. This study evaluated a collaborative swarm intelligence approach optimization algorithm, named the particle bee algorithm (PBA). The PBA is based on a particular aspect of bird (particle swarm optimization, PSO) and honeybee swarm (bee algorithm, BA) behaviors that integrates their advantages and proposes a self-parameter-updating technique to prevent being trapped into a local optimum in high dimensional problems. This study compares the performance of PBA with that of differential evolution (DE), evolutionary algorithms (EA), particle swarm optimization (PSO) and bee algorithm (BA) for multi-dimensional numeric problems. For test problems carried out in this work, colony sizes ranging from 75 to 100 of PBA can provide an acceptable convergence speed for an optimization search. Besides, elite and best bee PSO iteration sizes of (15, 9) to (30, 18) can provide an acceptable convergence speed for an optimization search. Results show PBA performance to be comparable to that of mentioned algorithms, and the potential for its being efficiently employed to solve benchmark numerical problems with high dimensionality.